The goal of this international team project is the development of a simulated testing environment based on an already existing open-source driving simulator for autonomous driving AIs. The key features are the following: An easy to use scenario Editor to build simulate short traffic situations to verify if the AI under test is able to safely handle them. A tool which is able to record and process the sensor data which is produced by the simulated cars sensors during the test scenarios. A rudimentary driving AI which can be customized to do several dangerous or faulty behaviors which will be used to check wether sets of test cases are able to detect the dangerous or faulty behavior.
- 4 Students from Mannheim
- 4 Students from Cluj
- Duration: 6 Month
- 2 Collaboration Weeks
- German students will visit the team in Cluj for 1 week (Paid by the University).
- Romanian students will visit the team in Germany for 1 week
- Development of a testing environment for autonomous driving
- Develop a scenario editor based on the already existing open source driving simulation Carla
- Development of a suitable Data Format and the necessary interfaces to the relevant sensors auf the simulated car to record all relevant data from the environment.
- Development of a test scenario replay chain where existing sets test cases can be executed evaluated automatically
- Development of a customizable rudimentary driving AI. The Goal is to build an AI which is capable of doing driving mistakes or dangerous behavior which can be defined by the user to verify if the test cases are able to find these mistakes or behaviors.
- It is strongly recommended to have some sort of programming skills. This is a very programming focused project so please don’t apply if you don’t have good programming skills.
- Experience with Python
- Experience with 3d Programming Engines like Unity or Unreal Engine
- Experience with Git-based development workflows
- Experience with artificial intelligence and reinforcement learning